One database per workspace
Each workspace has its own application database. No shared tables, no cross-tenant queries.
IQ Chat sends a conversation to the provider behind the model a user picks, from the catalog your administrators enable. This page states that data path, the other services content can reach, and the controls around them plainly, so a security review can start from facts.
Four stops. Each one has an owner and a control. The services content can reach beyond the model call are listed under Model providers below.
Users arrive through SSO with Google, Microsoft, or Auth0, provisioned and deactivated by SCIM 2.0. Two-factor authentication with TOTP and backup codes sits on top.
Each workspace runs in its own application database. Provider keys, connector tokens, and credentials are encrypted before storage. Company context only enters when a user connects Microsoft 365 or Google Workspace for themselves.
The conversation goes to the provider behind the model the user picked, from the catalog administrators approved. If that model fails, its backup takes over before any output is shown: the backup model administrators assigned, or, for a natively integrated model with none assigned, the same model through OpenRouter.
Answers return to the workspace. Files are built in isolated, network-less sandboxes, kept for the retention window administrators set, and durably deleted when it ends.
Every control below is in the shipped platform today. Administrators configure them per workspace from the console.
Each workspace has its own application database. No shared tables, no cross-tenant queries.
Provider keys, connector tokens, and credentials are encrypted before they are stored.
Microsoft 365 and Google Workspace connectors stay off until each user authorizes them. Administrators can keep them off for everyone.
Administrators choose which models appear, set the workspace default, and can restrict models, features, connectors, and tools by role.
Admin actions, sign-ins, and user activity are logged. Each entry stores the hash of the one before it, so a gap or an edit shows. Administrators verify the chain in the console and export it with the hashes.
Retention windows from 30 days to 10 years, set per workspace, with durable deletion of stored files when the window ends.
Word, Excel, PowerPoint, PDF, and image deliverables are produced in isolated sandboxes with no network access.
API requests run under personal keys and follow the user's role and model allow-list, with token and cost reporting per request.
A model call goes to the provider behind the chosen model, or to its backup when that model fails. Administrators decide which models appear in the catalog, and can remove any of them. The rows below list the other outside services content can reach.
Certifications, retention terms, residency, and provider contracts are confirmed for each deployment before it goes live.
A workspace on the IQNOVO platform, on your subdomain or your own domain, with its own application database and its own retention window.
The full stack deployed for one organization and scoped per engagement, for compliance teams that will not sign off on shared infrastructure.
No. IQNOVO does not train models and does not use workspace content to train or tune any model. Whether a model provider may use content depends on that provider's terms for the enabled models, which we confirm in writing for your configuration before rollout.
Users see their own conversations, and the people a conversation is shared with see that one. Administrators see the audit log of admin actions, sign-ins, and user activity, and the usage and cost analytics by model, user, agent, and feature. An administrator can open a conversation from the admin console, and each such view is written to the audit log first. Outside the workspace, conversation content reaches the provider behind the chosen model and the services listed on this page.
Yes. Retention windows run from 30 days to 10 years and are set per workspace, with durable deletion of stored files when the window ends. Because each workspace runs in its own application database, removal at the end of an engagement is confirmed in writing as part of the statement of work.
Yes. Send it through the contact form and a senior team member answers it. The spec sheet on the product page states what ships, and this page states the data path, so most questions can be checked against the site before we talk.
Certifications, retention terms, residency, and provider contracts are confirmed for each deployment, in writing, before it goes live. Ask in the security review and you get a plain answer, including what is not in place yet.
Yes. The dedicated deployment puts the full stack in place for one organization, scoped per engagement, for compliance teams that will not sign off on a shared platform.
Share your security review, your data-handling requirements, or the one question that decides it. A senior team member answers in writing, and says so if something is not in place.